skills.rest
human-centric-engineering
human-centric-engineeringOFFICIAL
Explore all skills by this author →
WebsiteGitHub

orchestration-knowledge-builder

Configure document ingestion, chunking, embeddings, and vector search for RAG-enabled agents.

Updated Jul 28, 2026
Search Tags:#embeddings#rag#knowledge base#agent orchestration#vector search#chunking#document ingestion
One-click install
npx skills add https://github.com/human-centric-engineering/resparkable --skill orchestration-knowledge-builder-human-centric-engineering
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: orchestration-knowledge-builder
Source: https://github.com/human-centric-engineering/resparkable/tree/main/.claude/skills/orchestration-knowledge-builder
Command: npx skills add https://github.com/human-centric-engineering/resparkable --skill orchestration-knowledge-builder-human-centric-engineering

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Agents without grounded knowledge hallucinate answers. This Skill sets up the full knowledge base lifecycle in the Resparkable orchestration system so agents answer questions from real documents instead of guessing. ## Core Features & Use Cases - Document Ingestion: Upload Markdown, text, CSV, EPUB, DOCX, and PDF files (up to 50 MB), with a two-step preview-then-confirm flow for PDFs and SSRF-protected URL fetching. - Chunking & Embeddings: Structural, semantic, and row-atomic CSV chunkers, plus explicit embedding backfill via the knowledge embed endpoint using providers like Voyage AI or OpenAI. - Agent Scoping & Retrieval Modes: Scope knowledge by category per agent and choose retrieval modes (model, first_turn, every_turn, keywords) with inline citation pills in chat. - Use Case: Upload your product documentation and FAQ files, generate embeddings, then bind the search_knowledge_base capability to a support agent so it answers customer questions with cited sources. ## Quick Start Set up a knowledge base by uploading my product docs, generating embeddings, and scoping them to my support agent.

Frequently Asked Questions about orchestration-knowledge-builder

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I set up a knowledge base for RAG agents?▼

Follow the 7-step process: configure an embedding provider, upload documents, trigger embedding generation via POST /knowledge/embed, tune search parameters, scope categories to the agent, pick a retrieval mode, and bind the search_knowledge_base capability.

What document formats are supported for knowledge base upload?▼

Supported formats are Markdown, plain text, CSV, EPUB, DOCX, and PDF, each up to 50 MB. PDFs require a two-step preview-then-confirm flow, while CSV uses row-atomic chunking with one chunk per data row.

Why does vector search return no results after uploading documents?▼

Embeddings are not generated automatically on upload. You must explicitly call POST /api/v1/admin/orchestration/knowledge/embed to backfill embeddings for all chunks before search returns results.

Can I use Anthropic as the embedding provider?▼

No, Anthropic does not offer embedding models. Use Voyage AI (free tier), OpenAI text-embedding-3-small or text-embedding-3-large, or a local option like Ollama via an OpenAI-compatible provider.

How do I limit which documents an agent can search?▼

Set the agent's knowledgeCategories array to specific category strings assigned at upload time. An empty array means the agent searches all categories with no filtering, which is often not intended.

Why is my hybrid search weight setting being ignored?▼

The settings JSON only honours the key bm25Weight, not keywordWeight. Storing keywordWeight is a silent no-op, so the resolver falls back to defaults and search behaves as vector-only.

Related Skills

Explore More Software Engineering Skills→
cocoindex-iococoindex-io
75

cocoindex

Design and run CocoIndex data transformation flows with incremental exports.

Official
Intermediate
thecloudtipsthecloudtips
18

Answer Synthesis

Synthesize sourced answers with inline citations and confidence ratings.

Community
Advanced
rayvoidxrayvoidx

rag-optimization

Optimize RAG pipeline retrieval, ranking, embedding, and vector-search parameters.

Community
Advanced
bilacchibilacchi

llm-ops

Implement RAG pipelines, vector databases, prompt engineering, and LLM cost evaluation for production AI systems.

Community
Intermediate